Blur Detection and Deblurring in Covid-19 Chest X-rays images using Generative Adversarial Networks
Shaheera A. Rashwan, Dina Abdelhafiz · 2022
Image blur happened always by patient’s organ motion and it is one of the reasons for image rejection in radiographic diagnostic imaging. More precisely, it causes the confusion in diagnosis. In our paper, we detect blur in Covid-19 Chest X-rays images using the variance of Laplacian focus measure. A certain threshold identifies if the image is blurry or not. We divide the image in 2x2 blocks to detect the blur in each block (patch) separately. Results show that the right and the left bottom block are always blurry in most Covid-19 chest X-rays images when we use a benchmark dataset of 1200 acquired images. Then, we deblur the blurry images using a deblurring technique based on deep learning to highly enhance the images. We compute the focus measure after deblurring and compare it with that before deblurring using subjective and objective evaluation techniques. Results show the efficiency of the deblurring technique to reduce the disturbance in the Covid-19 Chest X-rays images.